Skip to content

FL cobidders are descriptively distinct from other frequent losers

Intuition (plain-language)

Not every firm that loses a lot bids near a cartel. The cobidders differ descriptively from ordinary always-losers: they range across more buyers, concentrate on narrower product lines, and keep showing up beside the same winners. These are footprints of where these firms participate — descriptive economic content, not proof of a role played inside a scheme. Note too that the canonical label is not built from the frequent-loser flag (341 FL14 / 310 non-FL among the 651 positives), so any FL "enrichment" is a separate descriptive fact, not how the positives were chosen — and that enrichment does not survive matching on opportunity.

🟡 Among always-loser cobidders, adjudication-anchored cobidders look different from other always-losers along economically meaningful dimensions, with effect sizes (Cohen's d) in the 0.3–1.0 band. This is descriptive economic content about where these firms participate, not proof that any of them played a cartel role.

The canonical validation label is the 651 always-loser cobidders and is not FL-conditioned (composition 341 FL14 / 310 non-FL): the frequent-loser flag is never used to build the positive class. FL enrichment is therefore a separate descriptive fact — P(cobidder | FL14) = 12.5% vs 2.2% for non-FL always-losers (≈5.7×) — and it is not significant within matched opportunity strata (FL-enrichment permutation p = 0.067, ns). What survives is the structural participation profile below, restated as description.

  1. Buyer breadth (AN-008). Cobidders bid in 136.5 tenders on average vs 76.7 for non-cobidder FLs (d = +0.67); they cross 24.8 unique winners vs 13.5 (d = +1.00).
  2. Product concentration (AN-009). Cobidder product-portfolio HHI is 0.380 vs 0.288 for non-cobidder FLs (d = +0.39) — more focused product portfolios.
  3. Proximity to legal anchors (AN-008, AN-009). 1.5% of cobidders have a direct-defendant counterpart vs 0.2% of non-cobidder FLs (d = +0.46). FL-bidder-present markets have winner-side HHI 0.178 vs 0.303 — cobidders show up in the more contestable markets.
  4. Bid-level patterns (AN-010, AN-024). Consistent with, but not diagnostic of, credible losing roles — a pattern, not a proof.
  5. Survives volume matching (AN-041). Cobidders bid in ~1.8× more tenders, so the obvious worry is that the distinctness is just volume. Matching cobidders to non-cobidder FLs on tenders_count (standardized difference 0.49 → 0.00), the profile distinctness holds or strengthens: product HHI d = +0.47, winner-pair spread −0.56, median gap-to-winner −0.25. The distinctness is not a volume artifact. One honest casualty: the AN-031 bid-dispersion elevation collapses (+0.05, n.s.) and is dropped as a volume artifact.

These differences give the adjudication-anchored cobidder target economic content and make it harder to reduce the result to ordinary high-volume losing alone — a worry now directly addressed by the volume-matched audit. They describe a deployment profile; they do not identify a role.

Caveat. The distinctness that survives volume matching is structural (product specialization, winner spread, proximity to the winning bid). The bid-conduct distinctness is narrower: one channel survives (median gap-to-winner), while the bid-dispersion sub-signal (AN-041) and the bid-timing battery (AN-042: all Wilcoxon p ≥ 0.23) are documented nulls — so this is not a multi-channel behavioral signature. The reading stays 🟡 because the dimensions are observational, within-stratum, and single-source (BEC × CADE). The first-time-FL channel (AN-021) and the matched quadrant heterogeneity (AN-032) bound a causal-mechanism reading the finding does not assert, not the structural distinctness. The supporting hypothesis H:cobidder-profile-distinct is Partial (strongly supported) on the structural claim; promotion beyond 🟡 / to Confirmed needs non-BEC replication.

Sources.

  • Own analysis: AN-008 (FL characterization), AN-009 (network HHI + proximity), AN-021 (first-time channel — demoted), AN-024 (unified mechanism quadrants), AN-028 (standardized-diff battery, 7 dimensions × 3 comparisons), AN-031 (bid- level gap-to-winner d = -0.28 — behavioral distinctness beyond participation), AN-032 (matched quadrant heterogeneity audit — honest negative finding), AN-041 (volume-matched profile audit — structural distinctness survives matching on tenders_count; bid-dispersion sub-signal does not), AN-042 (volume-matched timing audit — candidate 2nd bid-conduct channel, null).
  • Cross-refs: H:cobidder-profile-distinct; docs/results.md.
  • Macros: \valBridgeTendCob (136.5), \valBridgeTendFLnc (76.7), \valBridgeUniqWinCob (24.8), \valBridgeHHICob (0.380), \valBridgeHHIFLnc (0.288), \valFLwinnerHHI (0.178), \valNonFLwinnerHHI (0.303), \valFTpsCoef (0.062), \valFTpsP (0.312).
  • Validation: backing scripts scripts/28_pbu_characterization.R, scripts/19_network_heterogeneity_2d.R, scripts/30_first_time_fl_matching.R, scripts/35_unified_mechanism.R.